When Parallels kicks off a round of SmartGuard snapshots, it does not prevent a large number of snapshots from firing concurrently. Parallels should add snapshot tasks to a queue and only process ONE snapshot at a time for any given storage device. This will reduce the total time needed to complete snapshots. This will decrease the total amount of disk space used during the process. If the process is disrupted by a machine or storage failure, then only one VM will be exposed to corruption. With the current approach, I've seen two dozen VMs light up at once, and the process takes upward of 7 minutes for snapshots stored on the fast internal iMac Pro drive. Likewise, for my machine with VMs on an iSCSI volume, the iSCSI target becomes unusable for everybody for the duration. I can't imagine how bad this Parallels SmartGuard experience must be for people on traditional (non-SSD) disks, where seek contention slows disk operations by orders of magnitude.